International Research Journal of Education and Technology

Forecasting stock prices is a crucial aspect of the financial sector, given the inherent volatility of the stock market. This task is complex due to the nonlinear, volatile, and dynamic nature of financial markets. It falls under the category of a time series problem, making conventional rules for predicting stock prices ineffective. Several methods exist for forecasting stock prices, including Logistic Regression, Support Vector Machines (SVM), Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN), Backpropagation, Naïve Bayes, and the ARIMA model. Among these, Long Short-Term Memory (LSTM) networks have emerged as the most suitable algorithm for time series analysis due to their ability to learn long-term dependencies in data. However, LSTM may lack robustness in feature selection and gradient boosting. The primary objective of this research is to forecast current market trends and improve the accuracy of stock price predictions. To address this, the study proposes the use of a hybrid model that combines LSTM with XGBoost, a powerful gradient-boosting algorithm known for its performance and flexibility in handling structured data. The integrated LSTM-XGBoost model leverages the strength of LSTM in time-series prediction and combines it with XGBoost’s ability to refine feature selection and reduce overfitting. This hybrid model is tested on stock market data for predictive analysis, demonstrating improved accuracy in terms of R² score, Mean Absolute Error (MAE), and predictive capability over traditional standalone models. By utilizing LSTM recurrent neural networks, the study achieved a prediction accuracy exceeding 95%, demonstrating the model's effectiveness in capturing the complexities of stock price movements.

Keywords-Stock Prediction, LSTM, XGBoost, Time-Series Analysis, ML, DL, Trade Open, Trade Close, Trade Low, Trade High, Financial Forecasting, Python.